CN110322491B - Algorithm for registering deformable mouse whole-body atlas and mouse image - Google Patents
Algorithm for registering deformable mouse whole-body atlas and mouse image Download PDFInfo
- Publication number
- CN110322491B CN110322491B CN201910501435.7A CN201910501435A CN110322491B CN 110322491 B CN110322491 B CN 110322491B CN 201910501435 A CN201910501435 A CN 201910501435A CN 110322491 B CN110322491 B CN 110322491B
- Authority
- CN
- China
- Prior art keywords
- atlas
- skin
- deformation
- mouse
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 210000003491 skin Anatomy 0.000 claims abstract description 49
- 210000000056 organ Anatomy 0.000 claims abstract description 27
- 210000000988 bone and bone Anatomy 0.000 claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 19
- 210000004072 lung Anatomy 0.000 claims abstract description 12
- 230000037396 body weight Effects 0.000 claims abstract description 10
- 230000036544 posture Effects 0.000 claims abstract 11
- 241000699666 Mus <mouse, genus> Species 0.000 claims description 69
- 210000001835 viscera Anatomy 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000005259 measurement Methods 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 11
- 241000699670 Mus sp. Species 0.000 claims description 8
- 210000004556 brain Anatomy 0.000 claims description 7
- 238000006073 displacement reaction Methods 0.000 claims description 7
- 210000004394 hip joint Anatomy 0.000 claims description 7
- 210000000323 shoulder joint Anatomy 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 210000003414 extremity Anatomy 0.000 claims description 4
- 239000000126 substance Substances 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 claims description 3
- 210000002216 heart Anatomy 0.000 claims description 3
- 210000003734 kidney Anatomy 0.000 claims description 3
- 210000004185 liver Anatomy 0.000 claims description 3
- 238000011524 similarity measure Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 210000000952 spleen Anatomy 0.000 claims description 3
- 210000003109 clavicle Anatomy 0.000 claims description 2
- 210000000078 claw Anatomy 0.000 claims description 2
- 238000012417 linear regression Methods 0.000 claims description 2
- 210000000614 rib Anatomy 0.000 claims description 2
- 210000001991 scapula Anatomy 0.000 claims description 2
- 210000003625 skull Anatomy 0.000 claims description 2
- 210000001562 sternum Anatomy 0.000 claims description 2
- 210000004003 subcutaneous fat Anatomy 0.000 claims description 2
- 208000031648 Body Weight Changes Diseases 0.000 claims 3
- 230000004579 body weight change Effects 0.000 claims 3
- 238000004364 calculation method Methods 0.000 claims 1
- 241001465754 Metazoa Species 0.000 description 7
- 238000010191 image analysis Methods 0.000 description 7
- 238000011160 research Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 3
- 230000000877 morphologic effect Effects 0.000 description 3
- 210000003484 anatomy Anatomy 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 210000002310 elbow joint Anatomy 0.000 description 2
- 210000003194 forelimb Anatomy 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 210000000629 knee joint Anatomy 0.000 description 2
- 108090000623 proteins and genes Proteins 0.000 description 2
- 238000003325 tomography Methods 0.000 description 2
- 238000005481 NMR spectroscopy Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000002591 computed tomography Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 210000001161 mammalian embryo Anatomy 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 239000002547 new drug Substances 0.000 description 1
- 238000009206 nuclear medicine Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000002603 single-photon emission computed tomography Methods 0.000 description 1
- 210000002356 skeleton Anatomy 0.000 description 1
- 210000004872 soft tissue Anatomy 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention discloses an algorithm for registering a deformable mouse whole body atlas and a mouse image, which mainly utilizes the deformable characteristic of the atlas to register a target image and maps an organ region in the atlas to a target individual to realize the division of the target organ region. In the registration, the postures, the body lengths and the body weights of the skin and the bones of the atlas are automatically adjusted to be close to the target in advance. The other organs meeting the statistical change rule can estimate corresponding positions and shapes according to the use condition Gaussian model of the atlas skin and the skeleton; the remaining organs are mapped using thin plate spline deformation from the surrounding organs. And obtaining a whole body atlas after the shape adjustment, extracting skin, bones and lungs, filling the whole body atlas with real gray values, registering the filled organs with the target by using a three-dimensional image registration method, and guiding the atlas to finish final deformation to obtain a final result. The invention simultaneously solves the problems of body posture change and organ individual shape difference in mouse image registration, and has higher registration precision and robustness.
Description
Technical Field
The invention belongs to the technical field of the research of medical images of small animals, and relates to an algorithm for registering a deformable mouse whole body map and a mouse image, which particularly focuses on the registration of a deformable characteristic of the mouse whole body map and the mouse image, maps an organ region in the map into a target image, and further divides the target image to obtain main organs of the mouse.
Background
With the rapid development of medical imaging technology, the small animal imaging research plays an important role in the pre-clinical cancer research and the trial production of new drugs. The progress and popularization of the small animal images provide new requirements for related image analysis work, a large amount of image data needs to be processed in unit time, and manual processing has the defects of strong subjectivity, poor repeatability and the like. Therefore, an automatic and objective small animal image analysis technology is urgently needed to improve the accuracy and efficiency of small animal image analysis. In automatic image analysis, the digital anatomical atlas plays an important role, and through registration of the atlas and a target individual, anatomical reference is provided for the target individual, and anatomical positioning is provided for expression of focus and gene information. Mice are the most common experimental individuals in the preclinical study stage.
The invention discloses a mouse digital anatomical map, which comprises three main categories of an embryo map, a brain map and a whole body map, and relates to the mouse whole body map. The general map of the mouse is mainly applied to four aspects in image analysis: the organ regions of the individual images are divided and measured through registration; providing anatomical reference for functional images such as Positron Emission Computed Tomography (PET), Single Photon Emission Computed Tomography (SPECT), Fluorescence Molecular Tomography (FMT) and the like; providing anatomy correlation positioning for multiple image acquisition of the same individual, wherein the multiple image acquisition comprises two types of conditions of different image mode acquisition of the same individual and different time point acquisition of the same individual; with the accumulation of the whole body gene information of the mouse, the anatomical positioning can be provided for the gene expression information.
Although there are many applications of the mouse whole-body atlas in image analysis, in the registration study using the mouse whole-body atlas, the morphological difference between the atlas and the target individual limits the accuracy, robustness and automation degree of atlas registration. Morphological differences are manifested in two ways: changes in shape due to body posture; the morphological differences of organs caused by factors such as weight, age, and population are reflected in the size ratio and relative displacement of organs, and fat thickness.
In order to realize the registration of the mouse whole body atlas and the target individual, various registration algorithms are provided at home and abroad and mainly comprise nonlinear deformation registration, movable joint registration and statistical shape model registration. The non-linear deformation registration firstly registers a target individual by defining a calibration point, and then registers the target individual with a body surface curved surface and a three-dimensional image, but the non-linear deformation mode cannot describe the posture change of the whole body of the mouse, so that organ distortion is easy to occur. In order to solve the problem of body posture change, the method provides that the movable skeleton atlas of the mouse is used for registration, the posture of the atlas is adjusted through the rotation of a skeleton joint, and the internal organs are driven to deform in a nonlinear mode. In order to solve the problem of internal soft tissue organ distortion, Hongkai Wang et al propose that a statistical shape model is used to model main organs of the whole body of mice with different weights, ages, sexes and populations, a statistical shape model of the whole body of the mice is constructed, and the deformation rules trained by the statistical shape model are registered, so that the deformation mode of the internal organs is ensured to be from the deformation rules among training samples, the basic topological structure is unchanged, the morphological change of the internal organs can be more accurately described than the nonlinear deformation mode, and the distortion problem is avoided. However, the existing three methods can not simultaneously solve the problems of the body posture change and the organ individual shape difference of the mouse, so that a mouse whole body map which can simultaneously realize the posture change and the organ shape change is constructed by Hongkai Wang and the like, and the map can control the map deformation by changing parameters such as the bone posture, the body length, the weight and the like. The invention provides an algorithm for registering with a mouse image on the basis of the deformable mouse atlas.
Disclosure of Invention
The invention aims to provide an algorithm for registering a deformable mouse atlas and a mouse image, which mainly solves the technical problem that the atlas registration has higher precision and robustness by utilizing the deformable characteristic of the whole-body atlas of a mouse and solving the problems of body posture change and organ individual shape difference of the mouse in the registration process. The images suitable for the invention comprise various tomography medical image modes, such as CT images, nuclear magnetic resonance images, nuclear medicine images and the like.
The technical scheme of the invention is as follows:
an algorithm for registering a deformable mouse whole body atlas and a mouse image comprises the following steps:
first, an anatomical landmark is selected from the mouse image.
The anatomical calibration points need to select joint points of bones and central points of internal organs, the number of the selected calibration points is not required, but in order to determine the body posture of the mouse, at least one anatomical calibration point is respectively contained in the limbs of the mouse, and whether the rest calibration points are added or not can be determined by a user according to the actual application effect. The selection of the calibration point can be manually selected or detected by an automatic method.
And secondly, registering the atlas skin and the skeleton to the target individual according to the selected calibration points.
And filling atlas skin and skeleton into gray values similar to corresponding organs of the target individual, and using three-dimensional gray image registration. The deformation mode in the registration process uses cubic B-spline deformation:
wherein T (x) is the transformation relation before and after the registration of the corresponding point x, xkFor control points, defined by regular grid vertices, β3(x) Is a cubic B-spline polynomial, pkIs the displacement vector of the control point of the B-spline, sigma is the control point distance, NkIs a set of control points acting at point x.
Mutual Information (MI) is used as a similarity measure for image registration:
wherein, IFAnd IMRespectively representing fixed and moving images, LMAnd LFRespectively a set of intensity information selected at a certain interval in moving and fixed images, p being the joint probability density, pFAnd pMEdge probability densities for the fixed and moving images, respectively, the edge probability density being determined by a joint probability density p over f and m, respectively LMAnd LFThe value of the variable above. The joint probability density was estimated by B-spline park windows as follows:
wherein T (x) is the deformation sideFormula omegaFIs a fixed picture IMA field, | ΩFI is the number of voxels in the image, wFAnd wMB-spline park windows, sigma, for fixed and moving images respectivelyFAnd σMFor scaling factor, from LMAnd LFAre determined, these parameters are directly derived from the moving image IM(x) And a fixed image IF(x) Or directly by the user.
In the registration, mutual information is used as similarity measurement of image registration, in order to enable the registration of the skin posture and the bone posture of the image to be more accurate, anatomical calibration point information of the first step is added on the basis of the similarity measurement, and minimum distance information of corresponding calibration points in the two images is used as an auxiliary measurement index of the similarity measurement. Therefore, the registered atlas skin and skeleton are obtained based on the anatomical calibration point by considering not only the image gray scale information but also the position relation of the known corresponding points.
And thirdly, adjusting the posture change of the body of the atlas according to the obtained skin and skeleton of the atlas which is registered based on the anatomical calibration point.
The Mouse Atlas of the Laboratory Mouse is adjusted for changes in posture using the means by which Hongkai Wang et al construct a deformable Mouse Atlas (journal article: ADeformable Atlas of the Laboratory Mouse). A posture control framework is defined in the atlas, the posture change of the skeleton of the atlas is controlled through a skeleton Subspace Deformation mode (SSD), the control framework is a control rod established for realizing model Deformation, and the control framework and the skeleton framework in the anatomical meaning are not the same concept. An external control frame is defined outside the atlas, skin deformation at shoulder joints and hip joints of the atlas is controlled in a harmonic coordinate transformation mode, and the deformation of the rest part of skin is controlled by using an SSD.
Based on the registration result of the second step, calculating to obtain a rigid deformation matrix of each control section in the registration process through the atlas and the skeleton before and after registration, and controlling the deformation of the atlas and the skeleton curved surface in an SSD mode through the rigid deformation of each control section, wherein the deformation mode is as follows:
p′i=(∑jωi,jTj)pi (4)
wherein p isiIs the four-dimensional homogeneous coordinate (x) of the ith vertex in the mouse atlasi,yi,zi,1),Tj4 x 4 homogeneous transformation matrix, omega, for controlling the rigid deformation of the jth control section of the skeletoni,jThe weight information is defined by the following formula for the influence weight of the ith vertex of the control segment in the map:
wherein D isi,jIs the shortest distance from the ith vertex to the jth control segment, SiIs a set of points with anatomical control over vertex i. If the vertex i is the vertex in the skull, limb, claw or sternum, SiI.e. the set of points of the skeleton to which the vertex i belongs. If the vertex i belongs to a vertex in the spine, ribs, scapula or clavicle, SiMiddle omegai,jPortions > 0 will comprise a multi-segmented skeletal structure. The weight information needs to use omegai,j/∑ωi,jNormalized to satisfy Σ ωi,j=1。
If the skin at the shoulder joint and the hip joint in the whole body atlas of the mouse is deformed by directly using an SSD, the curved surface collapse phenomenon easily occurs, the skin needs to be deformed by using a harmonic coordinate transformation mode, a simple external control frame is defined outside the atlas, the vertex of the simple frame is used as a control point, the deformation of the corresponding curved surface of the skin is controlled by the influence weight of the vertex on the curved surface of the skin, and the deformation is defined as follows:
wherein the content of the first and second substances,is the displacement vector of the jth vertex in the control frame,is the displacement vector of the ith vertex on the curved surface of the skin, hi,jIs the weight generated by the ith vertex on the curved surface of the skin by the jth vertex in the control frame calculated by the harmonic coordinates. And skin curved surfaces at the shoulder joint and the hip joint in the atlas are deformed by using the harmonic coordinate transformation, and the rest skin curved surfaces are deformed by using the SSD.
And fourthly, adjusting the body length and the weight of the atlas.
The change of the body length of the atlas is caused by the change of the length of the spine, and the deformation mode meets the linear scaling:
wherein, P is all vertex coordinates after the body length of the map is changed, P is0All vertex coordinates of the initial shape of the atlas, O is the extension of the central coordinate of the mouse atlas to P0The same dimension, can be subjected to matrix addition and subtraction operation, l0Is the spine length of the initial shape of the atlas, and l is the spine length of the atlas after the body length is changed.
The change of the body weight of the atlas is the change of the skin of the atlas caused by the accumulation of subcutaneous fat, and a skin deformation vector V caused by the body weightfThe learning method can be obtained by linear regression learning from samples, target individuals are standardized in advance when learning of weight change is carried out, and the samples have uniform body size and posture forms. Thus parameters for controlling the change in profile body weight need to be usedStandardization of wherein wkAnd lkFor the length and weight of the spine of the kth individual mouse, the change in coordinates of all vertices of the atlas due to weight change can be expressed as:
wherein, w0For initial profile body weight, the above formula assumes P and P0Having the same body length, the changes of all the vertex coordinates of the atlas caused by the changes of body length and body weight are reflected simultaneously by the following modes:
P=S(l,W(w,P0)) (9)
where l and w are two input variables that do not affect each other. In practice, the length and the weight of the mouse are changed simultaneously, and the change relationship of the length and the weight can be described by l ═ g (w), and can be obtained by statistics from training samples, and the change relationship can be used for describing the change of all vertex coordinates of the atlas caused by the change of the length and the weight of the atlas in the following way:
P=S(g(w),W(w,P0)) (10)
and fifthly, mapping the internal organs of the rest mice through the skin and the bones of the atlas after the posture, the body length and the weight are changed.
Statistical Shape Model (SSM) SSM in atlas is attached to skin and skeleton1Lung, heart, liver, spleen and kidney are subject to statistical shape model SSM2Two statistical shape model shape coefficients b1And b2The method is respectively subject to Gaussian distribution, and statistical correlation exists between the Gaussian distribution and the Gaussian distribution, and can be described by a Conditional Gaussian Model (CGM):
∑2|1=∑2+∑2,1(∑1)-1∑1,2 (13)
wherein the content of the first and second substances,is the mean, Σ, of a conditional probability distribution2|1Is a covariance matrix of the conditional probability distribution,sum Σ1Is b is1The mean and covariance matrices of (a) and (b),sum Σ2Is b is2Of the mean and covariance matrices, Σ2,1Sum Σ1,2Is b1And b2Cross covariance matrix between.∑1,∑2,∑2,1Sum Σ1,2The value of (c) can be found from a set of training samples. The formula according to the conditional gaussian model gives an estimate of the shape and location of the low-contrast organ using skin and bone. For the brain which does not meet the statistical deformation rule, control points are selected according to the skin and the skeleton near the brain, and the brain is mapped by using a deformation mode of a Thin Plate Spline (TPS).
And sixthly, registering the skeleton, the skin and the lung of the image spectrum with the target individual in a gray image registration mode.
And (4) obtaining the whole body map of the mouse after the posture, the length and the weight are changed after the fourth step, wherein the whole body map comprises all internal organs. Extracting skin, bone and lung from the changed atlas, filling the grey scale value similar to the corresponding organ of the target image, and carrying out registration on the three-dimensional grey scale image in the same second registration mode. Cubic B-splines are used as image deformation modes, and mutual information is used as registration similarity measures. And (4) obtaining a deformation field for three-dimensional gray image registration by registration, and controlling the map deformation after the shape posture, the body length and the weight are changed by using the deformation field to obtain a final registration result.
And seventhly, if the difference between the result obtained in the sixth step and the target individual is large, extracting skin and bones on the basis of the atlas deformation result in the sixth step, and repeating the processes from the fourth step to the sixth step until a registration result meeting the requirement is obtained.
The invention has the beneficial effects that: the deformable characteristic of the whole body atlas of the mouse is utilized, the body posture, the body length and the body weight of the atlas are adjusted in advance according to the target individual before image registration, and then image registration is carried out on the atlas and the target individual. Meanwhile, the problems of body posture change and organ individual shape difference in the existing mouse whole body atlas registration technology are solved, and the algorithm has higher registration precision and robustness. The invention has positive promotion effect on the small animal image analysis research, improves the data analysis capability of the small animal research before clinic and promotes the development of related biomedical research.
Drawings
FIG. 1 is a flowchart of an algorithm for registration of a whole-body atlas of a deformable mouse with an image of the mouse according to the present invention.
In the figure: (a) CT image of the target mouse; (b) (ii) whole body mapping of the transformable mouse; (c) target individual anatomical landmarks; (d) map anatomical landmarks; (e) skin and bone of the map after posture, length and weight changes; (f) a general body profile after posture, length and weight changes; (g) atlas skeleton, skin and lung fill images; (h) and (5) registering the result.
Detailed Description
The following detailed description of the invention refers to the accompanying drawings.
As shown in fig. 1, an algorithm for registration of a deformable mouse whole body atlas and a mouse image, wherein a target image is a CT image, and the deformable mouse whole body atlas is used for registration. Mainly comprises three parts: pre-adjusting the shape and posture of the skin and the skeleton of the whole body map of the deformable mouse; mapping the remaining internal organs according to the deformed skin and bone; and registering the deformed whole body atlas of the mouse to the target individual through three-dimensional gray level image registration. The specific implementation mode is as follows:
first, an anatomical landmark is selected from the mouse image. The target individual anatomical calibration points (c) comprise 7 joint points of a nasal tip, a cervical vertebra top, a coccyx top, a left forelimb elbow joint, a right forelimb elbow joint, a left hind limb knee joint and a right hind limb knee joint, and the invention is not limited by the selection of the 7 calibrations. The selection method can be selected from the target mouse CT image (a) or the bone segmentation result in a manual mode or an automatic detection mode. Map anatomy landmarks (d) can be selected from the joint points defined in the transformable mouse whole body map (a).
And secondly, registering the atlas skin and the skeleton to the target individual according to the selected calibration points. And filling atlas skin and skeleton into gray values similar to corresponding organs of the target individual, and using three-dimensional gray image registration. In the registration process, cubic B splines are used as an image deformation mode, and mutual information and the minimum distance information of corresponding calibration points are used as similarity measurement of image registration. Obtaining the registered atlas skin and skeleton based on the anatomical calibration points.
And thirdly, adjusting the posture change of the body of the atlas according to the skin and the skeleton of the atlas which is obtained in the second step and is registered based on the anatomical calibration points. And calculating to obtain a rigid deformation matrix of each control section in the control framework in the registration process through the atlas bones before and after registration, and controlling the deformation of the whole body atlas bone curved surface in an SSD mode through the rigid deformation of each control section.
If the skin at the shoulder joint and the hip joint in the whole body atlas of the mouse is directly deformed by using the SSD, curved surface collapse is easy to occur, the skin at the part needs to be deformed by using a harmonic coordinate transformation mode, a simple external control frame is defined outside the atlas, the vertex of the simple frame is used as a control point, and the deformation of the corresponding curved surface of the skin is controlled by the influence weight of the vertex on the curved surface of the skin. And the curved surface of the skin at the other parts is deformed by the SSD mode.
And fourthly, adjusting the body length and the weight of the atlas. And linearly adjusting the body length change of the atlas by using the length of the spine of the target individual. And calculating to obtain the weight change coefficient of the atlas by using the registration result of the skin of the atlas in the second step and the vector difference between the initial atlas and the deformation vector of the skin curved surface change caused by the weight change obtained by model training. Finally obtaining the skin and skeleton of the atlas after the posture, the length and the weight change (e).
And fifthly, mapping the internal organs of the rest mice through the skin and the skeleton of the atlas after the posture, the length and the weight are changed (e). The internal organs of the mice, lung, heart, liver, spleen and kidney, which satisfy the statistical change rule can be estimated by high-contrast organ skin and bones by using a conditional Gaussian model. The internal organs and brains of mice which do not meet the statistical change rule are mapped by using the deformation mode of the thin plate sample bands. Finally obtaining the whole body map (f) of the mouse after the posture, the length and the weight are changed, including all the internal organs of the mouse.
And sixthly, registering the skeleton, the skin and the lung of the image spectrum with the target individual in a gray image registration mode. And (c) obtaining a mouse whole body atlas (f) with changed posture, body length and weight after the fourth step, wherein the whole body atlas (f) comprises all internal organs, extracting skin, bones and lungs from the atlas after the change, filling the skin, bones and lungs by using gray values close to the organs corresponding to the target image to obtain a filling image (g) of the atlas skin, bones and lungs, and then registering the three-dimensional gray level image. In the registration process, cubic B splines are used as an image deformation mode, and mutual information is used as similarity measurement of image registration. And (4) obtaining a deformation field for three-dimensional gray image registration by registration, and controlling the deformation of the mouse whole body atlas (f) after the posture, the body length and the weight are changed by using the deformation field to obtain a final registration result (h).
And seventhly, if the difference between the result obtained in the sixth step and the target individual is large, extracting skin and bones on the basis of the atlas deformation result in the sixth step, and repeating the processes from the fourth step to the sixth step until a registration result meeting the requirement is obtained.
Claims (1)
1. An algorithm for registering a deformable mouse whole body atlas and a mouse image is characterized by comprising the following steps:
first, select anatomical index points from the mouse image
The anatomical calibration points need to select joint points of bones and central points of internal organs, in order to determine the body posture of the mouse, four limbs of the mouse respectively comprise at least one anatomical calibration point, and the addition of the rest calibration points is determined by a user according to the actual application effect;
secondly, registering the atlas skin and skeleton to the target individual according to the selected calibration points
Filling atlas skin and skeleton into gray values similar to corresponding organs of the target individual, and registering by using three-dimensional gray images; the deformation mode in the registration process uses cubic B-spline deformation:
wherein T (x) is the transformation relation before and after the registration of the corresponding point x, xkFor control points, defined by regular grid vertices, β3(x) Is a cubic B-spline polynomial, pkIs the displacement vector of the control point of the B-spline, sigma is the control point distance, NkIs a set of control points acting at the x point;
using mutual information as similarity measure for image registration:
wherein, IFAnd IMRespectively representing fixed and moving images, LMAnd LFRespectively a set of intensity information selected at a certain interval in moving and fixed images, p being the joint probability density, pFAnd pMEdge probability densities for the fixed and moving images, respectively, the edge probability density being determined by a joint probability density p over f and m, respectively LMAnd LFThe value of the above variable; the joint probability density was estimated by B-spline park windows as follows:
wherein T (x) is the deformation mode, ΩFIs a fixed picture IMA field, | ΩFI is the number of voxels in the image, wFAnd wMB-spline park windows, sigma, for fixed and moving images respectivelyFAnd σMFor scaling factor, from LMAnd LFAre determined, these parameters are directly derived from the moving image IM(x) And a fixed image IF(x) Or directly specified by the user;
mutual information is used as similarity measurement of image registration in registration, registration of image skin and bone postures is more accurate, anatomical calibration point information of the first step is added on the basis of similarity measurement, and minimum distance information of corresponding calibration points in two images is used as an auxiliary measurement index of the similarity measurement; registering atlas skin and skeleton, not only considering image gray information, but also considering the position relation of known corresponding points to obtain the registered atlas skin and skeleton based on anatomical calibration points;
thirdly, adjusting the body posture change of the atlas according to the obtained atlas skin and skeleton which are registered based on the anatomical calibration points
A posture control framework is defined in the map, the posture change of the map skeleton is controlled in a skeleton subspace deformation mode, and the control framework is a control rod established for realizing model deformation; an external control frame is defined outside the atlas, the skin deformation of shoulder joints and hip joints of the atlas is controlled in a harmonic coordinate transformation mode, and the deformation of the rest part of skin is controlled by using an SSD;
based on the registration result of the second step, calculating to obtain a rigid deformation matrix of each control section in the registration process through the atlas and the skeleton before and after registration, and controlling the deformation of the atlas and the skeleton curved surface in an SSD mode through the rigid deformation of each control section, wherein the deformation mode is as follows:
p′i=(∑jωi,jTj)pi (4)
wherein p isiIs the four-dimensional homogeneous coordinate (x) of the ith vertex in the mouse atlasi,yi,zi,1),Tj4 x 4 homogeneous transformation matrix, omega, for controlling the rigid deformation of the jth control section of the skeletoni,jFor the ith vertex of the control segment in the mapThe weight information is defined by the following formula:
wherein D isi,jIs the shortest distance from the ith vertex to the jth control segment, SiIs a set of points with anatomical control over vertex i; if the vertex i is the vertex in the skull, limb, claw or sternum, SiThe point set is the point set of the skeleton to which the vertex i belongs; if the vertex i belongs to a vertex in the spine, ribs, scapula or clavicle, SiMiddle omegai,jPortions > 0 will comprise multi-segmented skeletal structures; the weight information needs to use omegai,j/∑ωi,jNormalized to satisfy Σ ωi,j=1;
If the skin at the shoulder joint and the hip joint in the whole body atlas of the mouse is deformed by directly using an SSD, the curved surface collapse phenomenon easily occurs, the skin needs to be deformed by using a harmonic coordinate transformation mode, a simple external control frame is defined outside the atlas, the vertex of the simple frame is used as a control point, the deformation of the corresponding curved surface of the skin is controlled by the influence weight of the vertex on the curved surface of the skin, and the deformation is defined as follows:
wherein the content of the first and second substances,is the displacement vector of the jth vertex in the control frame,is the displacement vector of the ith vertex on the curved surface of the skin, hi,jThe weight of the ith vertex on the skin curved surface generated by the jth vertex in the control frame obtained by harmonic coordinate calculation; skin curve using place of shoulder joint and hip joint in atlasThe harmonic coordinate transformation is carried out for deformation, and the other skin curved surfaces are deformed in the SSD mode;
fourthly, adjusting the body length and the weight of the atlas
The change of the body length of the atlas is caused by the change of the length of the spine, and the deformation mode meets the linear scaling:
wherein, P is all vertex coordinates after the body length of the map is changed, P is0All vertex coordinates of the initial shape of the atlas, O is the extension of the central coordinate of the mouse atlas to P0Performing matrix addition and subtraction operation with the same dimensionality; l0The spine length of the initial shape of the atlas, and l is the spine length of the atlas after the body length is changed;
the change of the body weight of the atlas is the change of the skin of the atlas caused by the accumulation of subcutaneous fat, and a skin deformation vector V caused by the body weightfThe body weight change learning method is characterized in that the body weight change learning method is obtained through linear regression learning from samples, target individuals are standardized in advance when learning of body weight change is carried out, and the samples have uniform body size and posture forms; thus parameters for controlling the change in profile body weight need to be usedStandardization of wherein wkAnd lkFor the length and weight of the spine of the k-th individual mouse, the changes in all vertex coordinates of the atlas due to weight changes are expressed as:
wherein, w0For initial profile body weight, the above formula assumes P and P0Having the same body length, the changes of all the vertex coordinates of the atlas caused by the changes of body length and body weight are reflected simultaneously by the following modes:
P=S(l,W(w,P0)) (9)
wherein l and w are two input variables which do not influence each other; in practice, the length and the weight of the mouse are changed simultaneously, and the change relationship between the length and the weight is described by l ═ g (w), which is statistically obtained from the training sample, and the change relationship between the length and the weight of the atlas, which is caused by the change of the length and the weight of the atlas, of all the vertex coordinates can be described by the following way:
P=S(g(w),W(w,P0)) (10)
fifthly, mapping the internal organs of the rest mice through the skin and the bones of the atlas after the posture, the body length and the weight change
Statistical shape model SSM of skin and skeleton membership in atlas1Lung, heart, liver, spleen and kidney are subject to statistical shape model SSM2Two statistical shape model shape coefficients b1And b2Respectively obeying Gaussian distribution, and describing by using a conditional Gaussian model, wherein the statistical correlation exists between the two types of the Gaussian distribution:
∑2|1=∑2+∑2,1(∑1)-1∑1,2 (13)
wherein the content of the first and second substances,is the mean, Σ, of a conditional probability distribution2|1Is a covariance matrix of the conditional probability distribution,sum Σ1Is b is1The mean and covariance matrices of (a) and (b),sum Σ2Is b is2Of the mean and covariance matrices, Σ2,1Sum Σ1,2Is b1And b2Cross covariance matrix between;∑1,∑2,∑2,1sum Σ1,2The value of (a) is obtained from a training sample set; according to the formula of the conditional Gaussian model, the estimation of the shape and the position of the low-contrast organ can be given by using the skin and the skeleton; for the brain which does not meet the statistical deformation rule, selecting control points according to the skin and the skeleton near the brain, and mapping by using a deformation mode of a thin plate spline;
sixthly, registering the skeleton, the skin and the lung of the image spectrum with the target individual in a gray image registration mode
Obtaining a whole body map of the mouse after the posture, the body length and the weight change after the fourth step, wherein the whole body map comprises all internal organs; extracting skin, bones and lungs from the changed atlas, filling the skin, bones and lungs into gray values similar to the corresponding organs of the target image, and carrying out registration on the three-dimensional gray image in the second registration mode; using cubic B splines as an image deformation mode and using mutual information as registration similarity measurement; registering to obtain a deformation field for registering the three-dimensional gray level image, and controlling the map deformation after the shape posture, the body length and the weight are changed by using the deformation field to obtain a final registration result;
and seventhly, if the difference between the result obtained in the sixth step and the target individual is large, extracting skin and bones on the basis of the atlas deformation result in the sixth step, and repeating the processes from the fourth step to the sixth step until a registration result meeting the requirement is obtained.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910501435.7A CN110322491B (en) | 2019-06-11 | 2019-06-11 | Algorithm for registering deformable mouse whole-body atlas and mouse image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910501435.7A CN110322491B (en) | 2019-06-11 | 2019-06-11 | Algorithm for registering deformable mouse whole-body atlas and mouse image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110322491A CN110322491A (en) | 2019-10-11 |
CN110322491B true CN110322491B (en) | 2022-03-04 |
Family
ID=68119510
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910501435.7A Active CN110322491B (en) | 2019-06-11 | 2019-06-11 | Algorithm for registering deformable mouse whole-body atlas and mouse image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110322491B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113616273A (en) * | 2021-08-04 | 2021-11-09 | 长安大学 | Positioning block manufacturing method and system for precise replacement of artificial knee joint |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103514606A (en) * | 2013-10-14 | 2014-01-15 | 武汉大学 | Heterology remote sensing image registration method |
CN109242865A (en) * | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
Family Cites Families (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN100470587C (en) * | 2007-01-26 | 2009-03-18 | 清华大学 | Method for segmenting abdominal organ in medical image |
US9524552B2 (en) * | 2011-08-03 | 2016-12-20 | The Regents Of The University Of California | 2D/3D registration of a digital mouse atlas with X-ray projection images and optical camera photos |
CN104867104B (en) * | 2015-05-20 | 2017-12-15 | 天津大学 | Target mouse anatomical structure collection of illustrative plates acquisition methods based on the non-rigidity registration of XCT images |
CN106530338B (en) * | 2016-10-31 | 2019-02-05 | 武汉纺织大学 | MR image feature point matching process and system before and after biological tissue's non-linear deformation |
CN106920228B (en) * | 2017-01-19 | 2019-10-01 | 北京理工大学 | The method for registering and device of brain map and brain image |
CN108428245B (en) * | 2018-02-11 | 2022-03-08 | 中国科学院苏州生物医学工程技术研究所 | Slip image registration method based on self-adaptive regular term |
CN108053431A (en) * | 2018-02-24 | 2018-05-18 | 中原工学院 | A kind of non-rigid medical image registration method based on gradient distribution |
-
2019
- 2019-06-11 CN CN201910501435.7A patent/CN110322491B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103514606A (en) * | 2013-10-14 | 2014-01-15 | 武汉大学 | Heterology remote sensing image registration method |
CN109242865A (en) * | 2018-09-26 | 2019-01-18 | 上海联影智能医疗科技有限公司 | Medical image auto-partition system, method, apparatus and storage medium based on multichannel chromatogram |
Also Published As
Publication number | Publication date |
---|---|
CN110322491A (en) | 2019-10-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110335358B (en) | Personalized deformation method of deformable digital human anatomy model | |
Hill et al. | A strategy for automated multimodality image registration incorporating anatomical knowledge and imager characteristics | |
CN109215064B (en) | Medical image registration method based on superpixel guide | |
EP3286727B1 (en) | Whole body image registration method and method for analyzing images thereof | |
Wein et al. | Automatic bone detection and soft tissue aware ultrasound–CT registration for computer-aided orthopedic surgery | |
CN104867104B (en) | Target mouse anatomical structure collection of illustrative plates acquisition methods based on the non-rigidity registration of XCT images | |
CN110378881B (en) | Tumor positioning system based on deep learning | |
CN109509193B (en) | Liver CT atlas segmentation method and system based on high-precision registration | |
CN107680110B (en) | Inner ear three-dimensional level set segmentation method based on statistical shape model | |
CN114155286B (en) | Individualized registration method for anatomical morphology and material mechanics characteristic template library of skeleton CT image | |
CN106846330A (en) | Human liver's feature modeling and vascular pattern space normalizing method | |
CN112509022A (en) | Non-calibration object registration method for preoperative three-dimensional image and intraoperative perspective image | |
Spinczyk et al. | Automatic liver segmentation in computed tomography using general-purpose shape modeling methods | |
CN114792326A (en) | Surgical navigation point cloud segmentation and registration method based on structured light | |
CN115830016A (en) | Medical image registration model training method and equipment | |
CN115116586A (en) | Deformable statistical atlas construction method based on joint registration | |
Wang et al. | Deformable torso phantoms of Chinese adults for personalized anatomy modelling | |
CN110322491B (en) | Algorithm for registering deformable mouse whole-body atlas and mouse image | |
CN110570430A (en) | orbital bone tissue segmentation method based on body registration | |
WO2016072926A1 (en) | Whole body image registration method and method for analyzing images thereof | |
Otake et al. | Patient-specific skeletal muscle fiber modeling from structure tensor field of clinical CT images | |
Xie et al. | Tissue feature-based and segmented deformable image registration for improved modeling of shear movement of lungs | |
Urschler et al. | Assessing breathing motion by shape matching of lung and diaphragm surfaces | |
CN114359309A (en) | Medical image segmentation method based on index point detection and shape gray scale model matching | |
CN112598669B (en) | Lung lobe segmentation method based on digital human technology |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |